WCCC 2018: the 24th World Computer Chess Championship Article Accepted Version The WCCC 2018 report Krabbenbos, J., van den Herik, J. and Haworth, G. (2018) WCCC 2018: the 24th World Computer Chess Championship. ICGA Journal, 40 (3). pp. 182-193. ISSN 1389-6911 doi: https://doi.org/10.3233/ICG-190080 Available at http://centaur.reading.ac.uk/78436/ It is advisable to refer to the publisher’s version if you intend to cite from the work. See Guidance on citing . Published version at: https://doi.org/10.3233/ICG-190080 To link to this article DOI: http://dx.doi.org/10.3233/ICG-190080 Publisher: The International Computer Games Association All outputs in CentAUR are protected by Intellectual Property Rights law, including copyright law. Copyright and IPR is retained by the creators or other copyright holders. Terms and conditions for use of this material are defined in the End User Agreement . www.reading.ac.uk/centaur CentAUR Central Archive at the University of Reading Reading’s research outputs online WCCC 2018: The 24th World Computer Chess Championship Jan Krabbenbos, Jaap van den Herik and Guy Haworth1 Amersfoort, the Netherlands, Leiden, the Netherlands and Reading, UK The 24th World Computer Chess Championship started on July 16, 2018. Eight programs as in Table 1 participated in a round robin tournament of seven rounds. The tournament took place at the Stockholmsmässan in Stockholm, Sweden and was organised by the ICGA. The venue was part of the international conferences IJCAI, ECAI, AAMAS, ICML, ICCBR and SoCS. The main sponsor was Digital Game Technology, DGT. It was held in conjunction with the 8th World Computer Software Championship and the 2018 Computer Speed Chess Championship. The classical tempo was 1h45+15/move for each side. Table 1 The participants in WCCC 2018 (CPW, 2018) # Engine Author(s) State Operator Platform Cores Opening Book EGTs 1 BOOOT Alex Morozov UA A M 2 x Xeon 28 Arena Perfect, abs — 2 CHIRON Ubaldo Andrea Farina IT U A F 2 x AMD EPYC 7551 64 Private + GOI (polyglot) 6m Syzygy Frank Schneider, Kai K H and Timo 3 GRIDGINKGO DE 40 x Intel Xeon E5-2630 v3 320 Wolfgang Zugrav 6m Nalimov Himstedt and Rob Hyatt Haupt 4 JONNY Johannes Zwanzger DE J Z 108 x AMD Opteron 1296 Own 6m Syzygy Don Dailey, Mark Lefler, 6-man 5 KOMODO US Erdogen Günes 4 x Intel Xeon E7-8890 v2.2 58 Erdogan Günes and Larry Kaufman Syzygy LEELA CHESS Alexander Lyashuk, Gary Cyril Gyot (1-2), 8 V100 GPU (rounds 1-2) 6 CA — — — ZERO Linscott et al Mark Roddis (3-7) 2 x GTX 1080ti (rounds 3-7) 7 SHREDDER Stefan Meyer-Kahlen DE S M-K Intel Xeon Own, small, handmade Syz & Shredder 8 THE BARON Richard Pijl NL R P 2 x 14-core Xeon 28 Own 6m Syzygy This year the ICGA revived the tradition of a chess grandmaster explaining the games and entertaining the public. Evolution since the 1986 WCCC in Cologne has been considerable. In 1986 grandmaster Vlastimil Hort gave comments from the point of view of a chess player who knew better than the computer. In this tournament, the Swedish commentator Harry Schussler GM knew that almost all programs were stronger than he was. Nevertheless he did a very good job of pointing out which moves were typical for a computer and still very strong moves. He remarked that humans could learn a substantial amount from the computer manoeuvres. Round 1: LEELA CHESS ZERO – KOMODO 0-1, BOOOT – JONNY ½-½, GRIDGINKGO – THE BARON 1-0, CHIRON – SHREDDER ½-½ In the first round, LEELA CHESS ZERO and KOMODO played an exciting game that only failed to end in a draw because White overstepped the clock. In the opening, the key feature was the isolated central black pawn on d5. Was it a strong asset in the centre? Yes, but White, in taking the pawn would have 1 Corresponding author: [email protected] given Black even more chances. At the first opportunity, White preferred to castle kingside and not to take the pawn: indeed that was impossible. Only a few moves later small combinations played a key role in the game’s complexities. Black decided to play 15. … Na2, a very remarkable move, see Fig. 1a. The continuation was 15. … Na2 16. Rcd1 Qxb3 17. Bxd5 Rxd5 18. Qxd5 Qxd5 19. Rxd5 Be6 20. Rd2 resulting in an early endgame in which Black had two pieces for a rook and a pawn. The final phase, sometimes exciting and at other times rather boring, lasted more than 110 moves. At that time, the tablebases and human chess players with even mediocre knowledge of the game all knew the game was drawn. Even so, Black continued to play since White did not have the use of endgame tablebases and had to compute all variations. This took a lot of time and on move 134, White overstepped the time limit and sadly forfeited the draw. a b c Fig. 1. R1: a) LEELA CHESS ZERO – KOMODO 15b; b) BOOOT – JONNY 37b; c) GRIDGINKGO – THE BARON 34w. The game BOOOT – JONNY started as a quiet Semi-Slav. Black had a plan to attack the white king and for the execution of the plan JONNY transferred all its pieces to the kingside. BOOOT reacted adequately and gave Black the impression that the square f4 was overloaded since it should be used for the bishop as well as for the knight. For human beings this difficulty would be a challenging task to resolve but this is not so for computers of current playing strength. On move 37 as in Fig. 1b, Black played 37. … Bh2 in order to be able to play 38. … Nf4. To the end, the black bishop remained on h2. The black attack was vigorous but White was very alert in its defence, meanwhile creating counterplay in such a way that the game ended in repetitions of position. After a hard fight, this was a well-deserved draw for both sides. The second program that decided a game to its advantage was GRIDGINKGO. It won by accurate play and by taking very fine decisions. The main point was that it sacrificed the exchange for a passed pawn. It occurred in an attractive series of moves that started with 34. b4, see Fig. 1c. The continuation was 34. … Qb6 35. Qc2 Nxc5 36. dxc5. Black then decided to create space, sacrificing the e-pawn by playing 36. … e5. It was not clear if better moves were available. In the course of the game GRIDGINKGO was able to challenge THE BARON further on the kingside and succeeded in creating a passed f-pawn. A nice interplay of king, rook and pawns led to the finale, a well-deserved win. The game CHIRON – SHREDDER was a high-level game between opponents who had respect for the playing strength of the other. After a well-known opening, they soon followed their own paths. The strategic lines that were followed were understandable for human beings. No explanatory artificial intelligence was necessary to understand what motives were underlying the moves. The game was clear up to move 32 where Black decided to achieve a draw in the most decisive way although it was reached by a slightly surprising sequence of moves. Standings after round 1: 1-2 GRIDGINKGO, KOMODO 1; 3-6 BOOOT, CHIRON, JONNY, SHREDDER ½; 7-8 LEELA CHESS ZERO, THE BARON 0 Round 2: KOMODO – SHREDDER ½-½, THE BARON – CHIRON ½-½, JONNY – GRIDGINKGO ½-½, LEELA CHESS ZERO – BOOOT ½-½ The round resulted in four draws. One game, KOMODO – SHREDDER, was tough and inspiring, two games were very long and require some perseverance when being replayed, and the game between THE BARON and CHIRON ended in a draw without much excitement since they quietly exchanged all pieces. In the game KOMODO – SHREDDER, the most interesting point was after move 18. … axb6, see Fig. 2a, when White played 19. Nd6. The continuation was 19. … Bxd6 20. Bxb7 Rxe1+ 21. Rxe1 Rd8 22. Rd1 Bc7 23. Rxd8 Bxd8. After 24. Kf1 White had two small advantages, namely the bishop pair against bishop and knight and the distant file-a/b pawns whereas Black had pawns on files b and c. Nevertheless, the two advantages were insufficient for a win. THE BARON – CHIRON has been discussed above: a straightforward draw. JONNY – GRIDGINKGO promised to be a sharp game after 13. h4 Bxh4 14. Qf3, see Fig. 2b. However, Black managed to hold on to equality and around move 30 the pawn positions were so interwoven that both sides took 130 moves to find an opening. It was all in vain and on move 164, a draw was agreed when repetition of position occurred on the board after 5½ hours of play. a b c Fig. 2. R2: a) KOMODO – SHREDDER 19w; b) JONNY – GRIDGINKGO 13b; c) LEELA CHESS ZERO – BOOOT 16b. The game LEELA CHESS ZERO versus BOOOT was a game with one tense moment when White played 16. b5 as in Fig. 2c. After 16. … Na5 17. Nxd4 the weighing of the pros and cons involved very different features. Nevertheless, there was no imbalance and hence the game ended in a draw. Standings after round 2: 1-2 GRIDGINKGO, KOMODO 1½; 3-6 BOOOT, CHIRON, JONNY, SHREDDER 1; 7-8 LEELA CHESS ZERO, THE BARON ½ Round 3: BOOOT – KOMODO 0-1, GRIDGINKGO – LEELA CHESS ZERO 1-0, CHIRON – JONNY ½-½, SHREDDER – THE BARON 1-0 In this round, KOMODO played a very interesting game with BOOOT. It was a Berlin Wall opening: Black usually has a solid position but has to wait until the endgame before it is possible to have any advantage from the micro benefits collected to that point.